understanding bias and item missing data in...
TRANSCRIPT
Understanding Bias and Item Missing Data in NIBRS
American Society of Criminology 2017 Annual Meeting
Overcoming Measurement ChallengesNovember 17, 2017
Philadelphia, PA
Eman Abdu,Doug Salane and Peter Shenkin
Center for Cybercrime StudiesMathematics & Computer Science Dept.
John Jay College of Criminal JusticeCity University of New York
Acknowledgements
Many students have contributed: Boris Bonderenko, Raul Cabrera and Henry Gallo
Inter-university Consortium for Political and Social Research(ICPSR) and National Archive of Criminal Justice Data (NACJD)
FBI, Criminal Justice Information Services Division, UCR/NIBRS Groups
NSF, NASA and NIJ
Goals
Provide back ground on FBI’s National Incident-Based Reporting System (NIBRS)
Demonstrate utility of having NIBRS data in a relational data base (Oracle 12c)
Examine NIBRS data issues: nonresponse bias and extent of item missing data
Briefly discuss ongoing work
NIBRS Data Structure
• Group A offenses (53 crimes)– data on arrest, offense, offender, victim, property
– data on incident (administrative)
– 56 data elements in 6 main segments
• Group B offenses (11 crimes) – social crimes (victimless)– e.g., bad checks, disorderly conduct, driving under influence
– only recorded if there is an arrest
• new codes 2015: Identity theft (26F), Computer hacking (26G)
NIBRS Data Structure
• NIBRS Group A offenses – data in 6 major files or segments
• An incident can have multiple segments: victims, offenders, offenses, arrestees, property records
• Tied together by Agency Identifier (ORI) and incident number
• 13 Segment files 6 group A, 1 group B, 3 Windows files, 3 Batch Files
NIBRS Relational Database
• 59 Tables – 13 Segments + Codebook
• Enforces referential integrity – important when uploading new data
• Provides SQL query capability and processing capabilities (indices, partitioning, etc.)
• Extract required data and relationships
• Viewing and reporting tools
Sizes of NIBRS Segments
John Jay NIBRS Relational DatabaseSegment Type Record Counts
(in millions, first 7 rows)
’95-‘05 ’95-‘08 ‘95-’15
Columns (fields)
1.Administrative 29.1 44.1 79.7. 17
2.Offense 31.9 48.4 87.9 26
3.Property 33.3 50.7 93.8 25
4.Victim 31.7 48.2 88.0 55
5.Offender 32.9 50.0 90.8 12
6.Arrestee 8.0 12.4 23.9 21
7.Group B Arrest 9.9 14.6 26.5 19
8.Window Exceptional
Clearance
11,502 16, 611 38,357 27
9.Window Recovered
Property
7,086 11,074 18,952 35
10.Window Arrestee 156,791 179,559 241,187 32
Records per Segment in NIBRS
2015 2014 2010 2005 2000 1995
Administrative5,054,699 4,986,370 5,060,854 4,614,054 2,841,523 837,014
0ffense5,669,429 5,574,049 5,610,977 5,079,639 3,098,037 906,509
0ffender5,765,370 5,701,941 5,845,297 5,235,653 3,205,276 937,035
Victim5,677,586 5,587,973 5,636,428 5,067,759 3,075,362 889,743
Property 6.182,510 6,119,863 6,011,620 5,338,234 3,214,981 951,574
Arrestee1,671,621 1,667,262 1,606,460 1,334,625 769,630 227,090
Group B
Arrest
1,591,015 1,590,574 1,753,973 1,457,435 1,006,424 318,524
LEAs Reporting
6284 6258 5662 4862 3365 1255
LEAs Reporting at Least One Incident
Year Number % Increase Year Number % Increase
1995 1255 2006 4841 3.4
1996 1487 18.5 2007 4935 2.0
1997 1738 16.9 2008 5184 5.0
1998 2249 29.4 2009 5595 8.0
1999 2852 26.8 2010 5662 1.2
2000 3365 18.0 2011 5874 3.7
2001 3611 7.3 2012 6086 3.6
2002 3809 5.5 2013 6129 .7
2003 4287 12.5 2014 6258 2.1
2004 4525 5.6 2015 6284 .4
2005 4682 3.5
Sum of COUNT(*) Column Labels
NIBRS RELEASE
YEAR between 1 and 10
between 11
and 100
between 101
and 1,000
between 1,001
and 10,000
between
10,001 and
more than
100,000 Grand Total
1995 111 380 585 170 9 1,255
1996 128 422 712 211 14 1,487
1997 118 477 830 292 21 1,738
1998 158 598 1,114 356 23 2,249
1999 241 771 1,385 427 28 2,852
2000 304 884 1,624 516 37 3,365
2001 310 1,022 1,665 567 46 1 3,611
2002 383 1,042 1,716 616 51 1 3,809
2003 473 1,128 1,987 645 53 1 4,287
2004 504 1,237 2,019 705 58 2 4,525
2005 475 1,222 2,144 775 64 2 4,682
2006 488 1,233 2,245 807 66 2 4,841
2007 480 1,278 2,276 827 71 3 4,935
2008 476 1,381 2,396 859 70 2 5,184
2009 508 1,561 2,589 866 69 2 5,595
2010 484 1,624 2,615 871 66 2 5,662
2011 512 1,712 2,678 902 69 1 5,874
2012 541 1,754 2,794 928 68 1 6,086
2013 562 1,788 2,828 887 64 6,129
2014 532 1,929 2,858 874 65 6,258
2015 523 1,913 2,899 881 68 6,284
Code Tables in NIBRS (Type Criminal Activity)
CODE DESCRIPTION • B Buying/Receiving • C Cultivating/Manufacturing/Publishing • D Distributing/Selling • E Exploiting Children • J Juvenile Gang Involvement • G Other Gang • N None/Unknown Gang Involvement • O Operating/Promoting/Assisting • P Possessing/Concealing • T Transporting/Transmitting/Importing • U Using/Consuming • I Intentional Abuse and Torture
Code Tables in NIBRS (Victim Offender Relationship )
CODE DESCRIPTION VO Victim was Offender
NA Not applicable AQ Victim was Acquaintance
SE Victim was Spouse FR Victim was Friend
CS Victim Common-Law Spouse NE Victim was Neighbor
PA Victim was Parent BE Victim was Babysittee (the baby)
SB Victim was Sibling BG Victim was Boyfriend/Girlfriend
CH Victim was Child CF Victim was Child of Boyfriend / Girlfriend
GP Victim was Grandparent HR Homosexual Relationship
GC Victim was Grandchild XS Victim was Ex-Spouse
IL Victim was In-Law EE Victim was Employee
SP Victim was Stepparent ER Victim was Employer
SC Victim was Stepchild OK Victim was Otherwise Known
SS Victim was Stepsibling RU Relationship Unknown
OF Victim other family member ST Victim was Stranger
Code Tables in NIBRS(Bias Motivation)
• 11 Anti-White • 12 Anti-Black or African American • 13 Anti-American Indian or Alaska Native • 14 Anti-Asian • 15 Multi-Racial Group • 21 Anti-Jewish • 22 Anti-Catholic • 23 Anti-Protestant • 24 Anti-Islamic (Moslem) • 25 Other Religion • 26 Multi-Religious Group • 27 Atheism/Agnosticism • 31 Anti-Arab • 32 Anti-Hispanic or Latino • 33 Anti-Not Hispanic or Latino • 41 Anti-Male Homosexual (Gay)
•
• 42 Anti-Female Homosexual (Lesbian) 43 Anti-Lesbian, Gay, Bisexual, or Transgender, Mixed Group (LGBT) • 43 Anti-Lesbian, Gay, Bisexual, or Transgender, Mixed Group (LGBT) • 44 Anti-Heterosexual • 45 Anti-Bisexual • 51 Anti-Physical Disability • 52 Anti-Mental Disability • 88 None • 99 Unknown • 28 Anti-Mormon • 82 Anti-Other Christian • 84 Anti-Hindu • 85 Anti-Sikh • 61 Anti-Male • 62 Anti-Female • 71 Anti-Transgender • 72 Anti-Gender Non-Conforming • 16 Anti-Native Hawaiian or Other Pacific Islander
Entity Relationship(6 main segments)
Victim/Offender JoinORI
Code
Incident
Number
Offender
Sequence
No.
Offender
Age
Victim
Sequence
No.
Victim
Age
Incident
Date
1 CO0030400 CI0BRFRH-2 N 1 23 1 33 09-Nov-00
2 DE0020300 LT01KETVV0 N 0 00 1 39 16-DEC-02
3 DE0020600 LI01KVBRTU N 1 11 1 09 06-OCT-02
4 DE0020600 LI01KVBRTU N 1 11 2 08 06-OCT-02
5 DE0020600 LI01KVBRTU N 2 10 1 09 06-OCT-02
6 DE0020600 LI01KVBRTU N 2 10 2 08 06-OCT-02
7 DE0020600 LI01KVBRTU N 3 10 1 09 06-OCT-02
8 DE0020600 LI01KVBRTU N 3 10 2 08 06-OCT-02
9 DE0020600 LI01KVBRTU N 4 12 1 09 06-OCT-02
10 DE0020600 LI01KVBRTU N 4 12 2 08 06-OCT-02
11 IA0820200 7Z1C7REMQ-F 1 40 1 41 24-JAN-02
NIBRS Incidents with Multiple Segments (1995-2015)
Total Incidents 79,672,672
SegmentOne Two Three Four
Arrestee
17,329,233 21.75% 2,207,330 2.77% 423,080 0.53% 123,535 0.16%
Offender
71,715,271 90.01% 5,950,932 7.47% 1,320,391 1.66% 436,482 0.55%
Offense
72,083,712 90.47% 6,927,813 8.70% 596,652 0.75% 56,083 0.07%
Victim
73,380,728 92.10% 5,168,540 6.49% 746,749 0.94% 205,587 0.26%
NIBRS Incidents with Multiple Segments (2015)
Total Incidents 5,054,699
SegmentOne Two Three Four
Arrestee
1,259,886 24.93% 146,598 2.90% 24,349 0.48% 6,674 0.13%
Offender
4,532,042 89.66% 402,315 7.96% 80,674 1.60% 25,799 0.51%
Offense
4,504,537 89.12% 493,675 9.77% 49,541 0.98% 5,964 0.12%
Victim
4,585,143 90.71% 384,375 7.60% 56,808 1.12% 15,574 0.31%
NIBRS Release Year
Incidents where
Release year is not the
same as Incident year
Percentage of
incidents where
release year not the
same as incident year
Total Release year
Records
2015 68,091 1.35% 5,054,699
2014 67,010 1.34% 4,986,370
2013 72,288 1.43% 5,070,862
2012 68,357 1.30% 5,261,649
2011 63,905 1.26% 5,084,696
2010 61,940 1.22% 5,060,854
2009 60,658 1.20% 5,052,752
2008 56,882 1.13% 5,016,841
2007 58,303 1.17% 5,003,962
2006 59,110 1.20% 4,906,781
2005 52,351 1.13% 4,614,054
2004 46,690 1.14% 4,083,571
2003 39,856 1.10% 3,637,432
2002 36,941 1.07% 3,455,589
2001 200 0.01% 3,232,281
2000 22,407 0.79% 2,841,523
1999 20,484 0.95% 2,157,326
1998 102 0.01% 1,822,675
1997 56 0.00% 1,426,978
1996 81 0.01% 1,064,763
1995 168 0.02% 837,014
NIBRS Released Year vs. Incident Year
(1995 – 2015 data sets)
Study of selected offenses where offender used a computer
• Illustrates use of spreadsheet pivot tables to select desired data
• Requires data from the offender and offense segments
• Provides age and gender breakdown of the offenders
• Examine selected offenses where offender used a computer
Spreadsheet Pivot Tables
Offender Counts (Offender suspected of using a computer)Aggregated by Offense, Age and
Gender Year
Offense
Description Age Group Gender 2000 2001 2002 2003 2004 2005
Grand
Total
Embezzlement 11 – 20 F 11 7 5 8 8 14 53
M 4 5 6 8 6 7 36
20 – 30 F 17 18 19 22 20 29 125
M 11 13 14 14 12 23 87
31 – 40 F 9 9 18 20 13 31 100
M 8 9 12 7 12 13 61
41 – 50 F 5 7 7 6 8 21 54
M 3 8 4 4 4 10 33
51 – 60 F 2 4 1 1 4 12
M 1 4 2 3 10
Wire Fraud 11 – 20 F 1 3 3 4 2 2 15
M 9 9 9 13 13 12 65
20 – 30 F 1 6 3 6 14 16 46
M 7 12 18 22 27 22 108
31 – 40 F 3 2 8 9 8 30
M 4 8 11 12 13 21 69
41 – 50 F 1 5 3 3 3 6 21
M 4 2 2 8 4 5 25
51 – 60 F 1 3 1 2 4 11
M 2 2 2 1 1 6 14
Spreadsheet Pivot Tables
Offender Counts (Offender suspected of using a computer)
Aggregated by Offense, Age and Gender Year
Offense
Description Age Group Gender 2010 2011 2012 2013 2014 2015
Grand
Total
Embezzlement 11 – 20 F 17 15 8 22 35 41 138
M 10 13 12 25 27 23 110
20 – 30 F 42 47 47 82 58 83 359
M 31 45 35 67 64 75 317
31 – 40 F 31 40 36 60 72 53 292
M 29 23 26 24 35 38 175
41 – 50 F 24 25 29 28 32 35 173
M 8 16 12 21 16 26 99
51 – 60 F 12 8 11 13 8 18 70
M 4 10 3 12 8 9 46
Impersonation 11 – 20 F 24 17 23 45 30 39 178
M 23 25 46 119 56 47 316
20 – 30 F 58 73 99 110 121 123 584
M 110 78 109 112 129 153 691
31 – 40 F 60 57 61 110 100 128 516
M 52 57 84 111 112 129 545
41 – 50 F 33 44 51 61 55 61 305
M 31 41 54 53 81 76 336
51 – 60 F 13 12 14 29 23 26 117
M 19 19 31 38 32 43 182
BIAS due to Non Response
• Compare UCR and NIBRS reporting
• Examine Breakdown of Violent and Property Crimes in NIBRS and UCR
• Examine Larceny in NIBRS and UCR
NIBRS and UCRNIBRS
• 33 states certified, 38% report all crime in NIBRS
• Covers 30% of US population (96 million )
• 29% of all crime, 18 LEAs cover Group I cities
• 6648 LEAs participated in 2015, over 7000 in 2016
UCR
• 16,643 LEAs submitted data to UCR (18,439 total )
• Includes major municipalities, 83 LEAs covering Group I cities
• Mainly summary data but with some incident data
Breakdown of Violent Crimes
UCR Data and NIBRS
Crime Type UCR
(2014)
NIBRS
(2014)
UCR
(2015)
NIBRS
(2015)
NIBRS Data
(1995-2015)
Aggravated Assault 63.61% 62.29% 63.8% 62.44% 62.84%
Murder/Nonnegligent
Manslaughter1.22% 1.28% 1.30% 1.44% 1.17%
Rape (legacy definition) 7.21% 10.91% 7.50% 11.29% 10.04%
Robbery 27.96% 25.51% 27.30% 24.83% 25.96%
Increase in Violent Crimes
UCR and NIBRS (2013-2015)
Crime 2013 2014 2015
UCR NIBRS UCR NIBRS UCR NIBRS
murder 14,196 3,445 14,249 3,499 15,696 4,123
% increase .37% 1.57% 10.16% 17.83%
rape 79,770 28,855 84,041 29,723 90,185 32,279
% increase 5.35% 3.01% 7.31% 8.60%
robbery 341,031 73,354 325,802 69,512 327,374 70,923
% increase -4.47% -5.24% 0.48% 2.03%
aggravated
assault724,149 165,395 741,291 169,728 764,449 178,511
% increase 2.37% 2.62% 3.12% 5.17%
NIBRS Breakdown of Violent Crime
(1995 – 2015)
1995(1) – 2015(21)
Breakdown of Property Crimes
UCR and NIBRS
(2014 and 2015)
Crime Type UCR
(2014)
UCR % NIBRS
(2014)
NIBRS % UCR(2015)
UCR % NIBRS
(2015)
NIBRS %
Burglary 1,729,496 20.90% 486,554 20.24% 1,579,527 19.76% 461,674 19.44%
Larceny 5,858,496 70.77% 1,736,384 72.24% 5,706,346 71.39% 1,724,328 72.60%
Motor Vehicle
Theft689.527 8.33% 180,822 7.52% 707,758 8.85% 189,072 7.96%
Breakdown of Property Crimes
NIBRS
(1995 - 2015)
Crime Type NIBRS
(1995-2015) NIBRS %
Burglary 8,252,514 21.27%
Larceny 27,352,884 70.50%
Motor Vehicle
Theft3,192,197 8.23%
Breakdown of Property Crimes
NIBRS /UCR Trends
(2014 to 2015)
Crime Type
UCR NIBRS
Burglary -8.67% -5.11%
Larceny -2.60% -0.69%
Motor Vehicle
Theft2.64% 4.56%
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Larceny
Burglary
Motor Vehicle
Breakdown of Property Crime
NIBRS (1995-2015)
1995(1) – 2015(21)
Comparison of Larceny Details
UCR and NIBRS 2014 Data
(offense counts)
UCR Data NIBRS Data
Larceny Type Counts percentages Counts percentages
Pocket-picking 27,465 0.54% 6,884 0.40%
Purse-snatching 20,660 0.40% 5,653 .33%
shoplifting 1,097,444 21.47% 378,153 21.78%
From motor vehicles
(except accessories)
1,172,876 22.95% 358,120 20.62%
Motor vehicle
accessories
359,490 7.03% 79,794 4.60%
bicycles 184,575
3.61% 0 0%
From buildings 626,572
12.26% 225,598 12.98%
From coin-operated
machines
11,728 .23% 3970 .23%
All others 1,610,734 31.51% 678,212 39.06%
Totals 5,111,544 100% 1,736,384 100%
Comparison of Larceny Details
UCR and NIBRS 2015 Data
(offense counts)
UCR Data NIBRS Data
Larceny Type Counts percentages Counts percentages
Pocket-picking 28,532 0.5% 7,079 0.41%
Purse-snatching 22,825 0.4% 5,433 .32%
shoplifting 1,273,656 22.32% 390,971 22.67%
From motor vehicles
(except accessories)
1,370,664 24.02% 372,031 21.58%
Motor vehicle accessories 399,444 7.0% 77,014 4.47%
bicycles 205,428
3.6% 0 0%
From buildings 663,648
11.63% 214,311 12.43%
From coin-operated
machines
11,413 .2% 3,804 .22%
All others 1,730,735 30.33% 653,685 37.91%
Totals 5,706,345 100% 1,724,328 100%
Item Missing Data
• NIBRS has 53 data elements most of which are mandatory
• Data elements such as demographics of victim and offenders, relationships victim/offender and others are of interest to researchers and policy makers
• Compare rates of missing data in NIBRS and other sources such as SHR
• Examine item missing data in murders
NIBRS Unknown Murder Victim Information
(1995-2015)
victims Unknown age Unknown race Unknown gender
1995 458 6 1.31% 6 1.31% 0 0.00%
1996 643 13 2.02% 7 1.09% 3 0.47%
1997 749 18 2.40% 10 1.34% 0 0.00%
1998 975 39 4.00% 21 2.15% 7 0.72%
1999 1230 34 2.7% 27 2.20% 6 0.49%
2000 1695 82 4.84% 52 3.07% 17 1.00%
2001 1958 85 4.34% 49 2.50% 15 0.77%
2002 2053 95 4.63% 53 2.58% 15 0.73%
2003 2132 65 3.05% 52 2.44% 7 0.33%
2004 2358 104 4.41% 58 2.46% 21 0.89%
2005 3320 122 3.67% 76 2.29% 13 0.39%
2006 3404 111 3.26% 66 1.94% 25 0.73%
2007 3420 97 2.84% 62 1.81% 16 0.47%
2008 3252 97 2.98% 93 2.86% 28 0.86%
2009 3457 79 2.29% 54 1.56% 8 0.23%
2010 3430 46 1.34% 49 1.43% 9 0.26%
2011 3544 47 1.33% 77 2.17% 13 0.37%
2012 3689 52 1.41% 62 1.68% 11 0.30%
2013 3551 57 1.61% 57 1.61% 14 0.39%
2014 3596 49 1.36% 73 2.03% 23 0.64%
2015 4234 58 1.37% 71 1.68% 14 0.33%
NIBRS Unknown Offender Information1
(1995-2015)
Victims Offender
missing
unknown
demographics
unknown
age
unknown
race
unknown
gender
1995 458 4.37% 7.64% 6.99% 5.68% 4.80%
1996 643 7.93% 7.62% 7.00% 6.69% 5.29%
1997 749 10.41% 9.35% 8.14% 7.21% 6.28%
1998 975 7.08% 9.85% 8.82% 6.77% 5.23%
1999 1230 9.02% 9.27% 7.97% 7.64% 5.93%
2000 1695 9.44% 15.16% 14.40% 10.86% 9.79%
2001 1958 11.90% 11.64% 10.73% 8.27% 7.46%
2002 2053 10.23% 12.96% 11.69% 8.91% 7.60%
2003 2132 11.30% 12.24% 10.79% 9.29% 7.88%
2004 2358 10.69% 15.18% 13.02% 11.28% 9.16%
2005 3320 11.20% 19.94% 17.95% 14.46% 12.02%
2006 3404 11.72% 18.51% 16.69% 12.66% 11.05%
2007 3420 12.54% 15.26% 13.57% 9.30% 7.63%
2008 3252 13.47% 14.94% 12.67% 10.61% 8.30%
2009 3457 12.09% 15.33% 13.51% 9.98% 7.84%
2010 3430 13.29% 14.46% 13.27% 9.04% 7.49%
2011 3544 12.39% 15.77% 14.11% 10.38% 8.94%
2012 3689 13.53% 15.83% 14.10% 10.11% 8.65%
2013 3551 12.56% 14.81% 13.38% 9.63% 8.39%
2014 3596 11.43% 14.35% 12.26% 10.65% 8.79%
2015 4234 13.51% 14.65% 14.27% 11.45% 9.54%1The unit of analysis is victims.
Ongoing Work
• Time series studies to examine NIBRS missing data, victim-offender relationships, circumstances, location and weapon used
• Extract data for specific studies and make it available in Excel Pivot Tables or Data Cubes
• Examine effects of police reporting practices on the data, e.g., inaccurate incident times
• Prepare for additional NIBRS reporting. DOJ, OJP, BJS and FBI program to create a nationally representative crime sample and NIBRS compliant operational systems increasing NIBRS reporting. (Mainly an IT effort)
• Make the relational database publicly available through use of the Oracle Data Pump utility
Thank You
Eman Abdu
Doug Salane and Peter Shenkin
212 237-8836
Center for Cybercrime Studies
Math & CS Dept.
John Jay College of Criminal Justice